AIDS and Behavior

, Volume 12, Issue 4, pp 594–603

The Relationship Between Alcohol Consumption and Unprotected Sex Among Known HIV-discordant Couples in Rwanda and Zambia

Authors

  • Matthew E. Coldiron
    • Rwanda Zambia HIV Research Group, Rollins School of Public HealthEmory University
    • Hubert Department of Global Health, Rollins School of Public HealthEmory University
  • Elwyn Chomba
    • University Teaching Hospital School of MedicineUniversity of Zambia
  • Cheswa Vwalika
    • Rwanda Zambia HIV Research Group, Rollins School of Public HealthEmory University
    • Zambia Emory HIV Research Project
  • Etienne Karita
    • Rwanda Zambia HIV Research Group, Rollins School of Public HealthEmory University
    • Project San Francisco
  • Kayitesi Kayitenkore
    • Rwanda Zambia HIV Research Group, Rollins School of Public HealthEmory University
    • Project San Francisco
  • Amanda Tichacek
    • Rwanda Zambia HIV Research Group, Rollins School of Public HealthEmory University
    • Hubert Department of Global Health, Rollins School of Public HealthEmory University
  • Leia Isanhart
    • Rwanda Zambia HIV Research Group, Rollins School of Public HealthEmory University
  • Susan Allen
    • Rwanda Zambia HIV Research Group, Rollins School of Public HealthEmory University
    • Hubert Department of Global Health, Rollins School of Public HealthEmory University
  • Alan Haworth
    • The Ministry of Health Counseling Services Unit, Chainama Hills Hospital & Department of PsychiatryUniversity of Zambia School of Medicine
Original Paper

DOI: 10.1007/s10461-007-9304-x

Cite this article as:
Coldiron, M.E., Stephenson, R., Chomba, E. et al. AIDS Behav (2008) 12: 594. doi:10.1007/s10461-007-9304-x

Abstract

Although alcohol abuse is highly prevalent in many countries in sub-Saharan Africa, little is known about the relationship between alcohol consumption and risky sexual behavior in these settings. An understanding of this relationship is particularly important given the high prevalence of HIV that exists in many of these countries. This study analyzes data collected from members of cohabiting HIV-discordant couples regarding alcohol consumption and self-reported condom use. After controlling for demographic and socioeconomic co-factors, alcohol use by male partners of HIV-discordant couples was associated with self-reported unprotected sex at follow-up. Counseling about alcohol use should be part of HIV testing and counseling programs, particularly among those found to be HIV-positive.

Keywords

HIVAlcoholCouplesCondoms

Introduction

Sub-Saharan Africa is disproportionately affected by the HIV/AIDS epidemic, with 24.7 million infected persons, and 2.9 million of the total global AIDS deaths in 2006 (UNAIDS/WHO 2006). It is well known that the predominant mode of transmission of HIV in sub-Saharan Africa is heterosexual sex (Quinn et al. 2000). HIV-negative persons in stable partnerships with HIV-positive individuals are at extremely increased risk of acquiring HIV (Hugonnet et al. 2002; Allen et al. 1992), making HIV-discordant couples primary targets for the prevention of the spread of HIV disease in Africa (Allen et al. 1992, 2003).

Alcohol use is high in most African countries, and alcohol use disorders are a common problem (Lopez and Matthers 2006). Among the 20 African countries identified by the WHO as having very high adult and child mortality (a list that includes both Rwanda and Zambia), it was estimated that the average adult consumes 7.1l of absolute alcohol per year (Rehm et al. 2004). Unlike Europe and the Americas, however, there is a relatively low percentage of adults who drink in these countries (55% of men and 30% of women), so when alcohol consumption is averaged out among drinkers, the annual consumption per drinker rises to 16.6l of absolute alcohol, the largest of any region in the world (Rehm et al. 2004). The overall pattern of drinking in these 20 sub-Saharan African countries is ranked as the second-most detrimental in the world, indicative of high rates of binge drinking and alcohol dependence (Rehm et al. 2004).

There are data that suggest that alcohol directly increases biologic and physiologic susceptibility to infection with HIV, and there are also data that suggest that alcohol use promotes risky sexual behavior, which has been shown repeatedly to be a risk factor for contracting HIV (Jonsson et al. 1997; Radcliffe et al. 2001; Crosby et al. 2002). This is likely not an “either-or” question: both direct biological effects and indirect behavioral effects contribute to alcohol’s influence on HIV transmission. It is possible that alcohol has a direct effect on acquisition of HIV. Wang et al. (2002) showed that, in vitro, cells cultured with alcohol had significantly higher up-regulation of the CCR5 receptor, which is a major protein involved in attachment of HIV to the cell membrane. With a higher density of CCR5 expressed on cell membranes, HIV would have greater opportunity to enter the cell. Beyond its effects on disease transmission, there is also evidence that alcohol allows for increased HIV replication and HIV disease progression in those already infected with HIV (Crum et al. 1996; Cook et al. 1997; Fong et al. 1994; Bagasra et al. 1993). The simplest and most obvious explanation of the behavioral link between alcohol and HIV infection is that alcohol use leads to risky sexual behaviors, although evidence of this relationship from African countries is limited. Among Botswanan men, heavy alcohol use ( >21 drinks/week) was strongly associated with unprotected sex (OR 3.48), multiple partners (OR 3.08), and paying for sex (OR 3.65), and a dose–response relationship emerged when comparing the sexual practices of abstainers, moderate drinkers, and heavy drinkers (Weiser et al. 2006). Simbayi et al. (2004) found that alcohol use was associated with a greater numbers of sex partners, higher rates of unprotected intercourse and more frequent condom failures. Zablotska et al. (2006) found that individuals who reported consuming alcohol before sex were more likely (RR 1.38 among men, 1.28 among women) to report inconsistent condom use in the previous 6 months. Among a sample of eighth grade students in South Africa, those who reported ever using alcohol and/or marijuana were 2.2 times more likely to report inconsistent condom use, i.e. during 25–75% of sexual encounters, than those who were nonusers (Palen et al. 2006). Given alcohol’s potential role in the transmission of HIV and the high risk of transmission among members of HIV-discordant couples, a better understanding of alcohol’s effects on sexual behavior in these couples could provide important insight into slowing the rate of transmission of HIV in sub-Saharan Africa. The current analysis is unique in that it examines alcohol’s effects on condom usage among HIV-discordant couples in which both partners are aware of the couple’s sero-status. Couples in the study are exposed to behavioral change interventions to reduce heterosexual HIV transmission. Thus both partners are aware of the risk of HIV transmission and have information to aide them in managing this risk. In this context we examine how alcohol may influence the risk-taking process and lead to the non-use of condoms among sero-discordant couples.

Data and Methods

Procedures

Study participants were discordant cohabiting couples (one partner is HIV sero-positive while the other is sero-negative) who were identified during confidential HIV Couples’ VCT (CVCT) at sites operated by Project San Francisco in Kigali, Rwanda, and the Zambia-Emory HIV Research Project in Lusaka, Zambia, operating under the umbrella of the Rwanda Zambia HIV Research Group, Emory University, Atlanta, Georgia, US. The same-day couples’ VCT services have been described previously and included free treatment for syphilis, condom skills training, and free condoms (McKenna et al. 1997). Eligibility criteria for enrollment in the study included: (i) cohabitation in a sexual relationship for at least 6 months and residence in Kigali, Rwanda, or Lusaka, Zambia, at the time of enrollment; (ii) discordant HIV sero-status between members of the couple; and (iii) age 18–48 years for women and 18–65 years for men. Couples who had become concordant positive between CVCT screening and their enrollment into prospective follow-up were excluded. In total, 990 couples were enrolled in Rwanda and 947 couples in Zambia. Both partners signed a joint informed consent and were assigned a study number to ensure confidentiality. At enrollment, demographic and socioeconomic information, prior sexual history, and frequency of alcohol use was documented for both members of the couple. Questionnaires were conducted by trained research nurses at the clinical research sites, and were conducted in private and administered separately to men and women. At a 3 month follow-up visit, the number of encounters with and without a condom, both with the spouse and with other partners, was recorded. Sexual exposures within the couple were quantified by self-reporting of each member. Each member of the couple was interviewed separately to allow reporting of outside contacts, and to indicate privately whether there had been problems with condom use with the spouse. Discrepancies between husbands’ and wives’ reports (noted at <15% of visits) were resolved with a repeat interview of both partners, separately and then together, to achieve consensus. Additional counseling was provided on request and when unprotected contacts were reported. Data on alcohol consumption was collected using a revised version of the AUDIT tool (Alcohol Use Disorders Identification Test) (WHO 2001), which asks questions on frequency and quantity of drinking, impaired control due to drinking, feelings of guilt associated with drinking, and the occurrence of black-outs or alcohol related injuries. In Zambia, it is common place for men to consume alcohol from a shared tin; hence the AUDIT tool was revised to reflect these drinking practices, and interviewers were trained to calculate approximate quantities of alcohol consumed through asking questions on the size of the shared tin and the amount of alcohol consumed by each individual.

Measures

HIV status was established with three rapid antibody tests and ELISA confirmation in the case of any discrepant rapid test results. Three measures of alcohol consumption are used. The first alcohol measure was a binary “ever/never” drank during the previous year. The second and third were based on responses to a battery of six questions about frequency of alcohol behaviors that included; drinking >6 drinks on one occasion, not being able to stop drinking once started, failing to meet normal expectations because of drinking, needing an eye-opener, feeling guilty about drinking, and forgetting what happened the night before because of drinking. The second measure was a continuous score from 0 to 6 based on binary “ever/never” responses to these questions. The third measure was a composite of measured frequencies, e.g. weekly/daily, of the six behaviors above, on a scale of 6–36, with 6 representing “never” answers to all 6 questions, and 36 representing “almost daily” answers to the questions. The variable was analyzed as a categorical variable by dividing the distribution of the variable into thirds: the reference category was a score of 6, the intermediate category a score of 7–9, and the highest category was a score of ≥10. At 3 monthly follow-up visits when couples reported they had experienced unprotected sex, they were asked if it was the result of alcohol consumption. However, in order to reduce the potential bias introduced by social desirability reporting–that is respondent’s under reporting of the link between alcohol and unprotected sex—we model independently collected measures of alcohol consumption and unprotected sex.

Analysis

The analysis focuses on the couple as the unit of analysis: with alcohol consumption reported by males, and condom use reported by both males and females. The analysis examines a binary outcome coded one if the couple reported at least one episode of unprotected sex at their first follow-up visit three months after enrollment, and the key covariate of interest was alcohol consumption by the man in the year prior to enrollment quantified using three separate measures of alcohol consumption. The analysis also controlled for other potential influences on unprotected sex, categorized as demographic (including age, age difference in the couple, number of years cohabiting, number of years living in Lusaka or Kigali, fertility desires and location where childhood was spent) and socioeconomic (including employment history, individual income, combined income, marriage status, home ownership, electricity in the home, religion, and language skills). The cohorts from Kigali and Lusaka were analyzed separately, and separate models that controlled for the demographic and socioeconomic cofactors were developed for each country. The first stage in the analysis was bivariate analysis to assess relationships between the unprotected sex and alcohol exposure variables and other couple characteristics, using Chi-squared and Fisher’s Exact tests to assess the statistical significance of compared proportions. In the second stage of analysis, multivariate analysis was performed. Separate logistic models were fitted for each country; the outcome in each was unprotected sex. Each measure of alcohol exposure was tested in the model, and each model controlled for a range of demographic and socio-economic variables. Interaction terms between the alcohol measure and the other independent variables were tested in each model. Data entry and management were performed with Microsoft Access, and analysis was performed with the SAS statistical package (version 9; Statistical Analysis Software, North Carolina, USA).

Results

Table 1 shows the distributions of variables considered in the analysis. In both Rwanda and Zambia, female participants were younger than male participants (Rwanda; χ = 8.52, df = 4, P < 0.05, Zambia χ = 6.65 df = 4, P < 0.05). While income was low for men and women, it was significantly lower among women, who were less likely to work regularly (Rwanda χ = 2.47, df = 3, P < 0.05, Zambia χ = 1.26, df = 3, P < 0.05). In both countries, religious affiliations were widely varied, though Catholics and Pentecostals were the largest groups represented. In Rwanda, 71% of males and 45% of females had used alcohol in the year prior to enrollment, and 23% of couples enrolled reported at least one episode of unprotected sex at their first follow-up visit after enrollment. In Zambia, 71% of men and 20% of women reported alcohol use in the year prior to enrollment, and 23% of couples enrolled reported at least one episode of unprotected sex at their first follow-up. Of note, 39% of couples enrolled in Zambia did not follow-up after their enrollment, compared to only 4% in Rwanda. Table 2 compares drinking patterns in Rwanda and Zambia. While they are generally similar, a significantly higher rate of binge drinking is seen in Zambia, which accounts for most of the differences seen in the composite alcohol consumption scores.
Table 1

Demographic, socioeconomic, and behavioral description of HIV-discordant cohorts in Rwanda and Zambia, 2003–2005

 

Rwanda

Zambia

Men

Women

Men

Women

N = 990

%

N = 990

%

χa

N = 947

%

N = 947

%

χa

Demographic

Age

    ≤25

93

9

313

32

 

72

8

343

36

 

    26–29

139

14

201

20

 

160

17

186

20

 

    30–33

204

21

217

22

 

212

22

179

19

 

    34–39

215

22

174

18

 

229

24

151

16

 

    ≥40

329

33

75

8

 

273

29

87

9

 

    Missing

10

1

10

1

8.52*

1

0

1

0

6.65*

Years cohabiting with partner in studyb

    0–2

254

26

   

240

25

   

    3–5

286

29

   

275

29

   

    6–9

223

23

   

166

18

   

    10+

227

23

   

266

28

   

Where participant lived until 16 years old

    City

177

18

175

18

 

454

48

555

59

 

    Town

87

9

71

7

 

253

27

236

25

 

    Village

726

73

744

75

1.50

240

25

156

16

9.21*

Socioeconomic

Worked for money in past year

    Throughout the year

518

53

247

25

 

507

54

306

32

 

    Seasonally/Part of the year

108

11

83

8

 

250

26

112

12

 

    Once in a while

235

24

102

10

 

54

6

45

5

 

    Unemployed

123

13

555

56

 

136

14

484

51

 

    Missing

6

0

3

0

2.47*

0

0

0

0

1.26*

Monthly pay (ZMK and FRW converted to 2007 USD)

    0

122

12

554

56

 

132

14

483

51

 

    $0 < $15

131

13

156

16

 

42

4

128

14

 

    $15–30

247

25

168

17

 

171

18

142

15

 

    $30–60

303

31

75

8

 

308

33

117

12

 

    >$60

186

19

36

4

 

294

31

77

8

 

    Missing

1

0

1

0

10.01*

0

0

0

0

7.95*

Religion

    Catholic

433

44

300

30

 

261

28

229

24

 

    Pentecostal

207

21

306

31

 

238

25

298

31

 

    Jehovah’s Witnesses

25

3

29

3

 

56

6

50

5

 

    Seventh Day Adventists

118

12

158

16

 

63

7

63

7

 

    Other

21

2

23

2

 

144

15

149

16

 

    Muslim

116

12

95

10

 

9

1

11

1

 

    United Church of Zambia

0

0

0

0

 

74

8

85

9

 

    None

46

5

28

3

 

56

6

16

2

 

    Baptist

7

1

13

1

 

32

3

32

3

 

    Anglican

17

2

38

4

5.35*

14

1

14

1

4.32*

Behavioral

Reported unprotected sex at 3-month followupc

    No

  

725

73

   

364

38

 

    Yes

  

227

23

   

216

23

 

    Missing

 

38

4

   

367

39

  

Reported alcohol use in year prior to enrollment

    No

287

29

545

55

 

278

29

753

80

 

    Yes

702

71

445

45

 

669

71

194

20

 

    Missing

1

0

0

0

9.21*

0

0

0

0

7.82*

aFisher’s Exact Test for comparison of values between men and women

bOne value reported for both members of the couple

cUnprotected sex reported by both members of couple, but only female reports used in analysis

P-value < 0.01

Table 2

Male alcohol consumption patterns, Rwanda and Zambia, 2003–2005

 

Rwanda (= 990)

Zambia (= 947)

n

%

n

%

In the last year, how often did you drink alcohol?

Never

287

29

278

29

Once a month or less

109

11

46

5

2–4 times a month

159

16

203

21

2–3 times a week

209

21

248

26

4 or more times a week

225

23

172

18

Missing

1

0

0

0

In the last year, have you had 6 or more drinks on more than one occasion?

Never

840

85

369

39

Less than monthly

72

7

93

10

Monthly

33

3

103

11

Weekly

27

3

267

28

Daily or almost daily

16

2

115

12

 

2

0

0

0

In the last year, have you found that you were not able to stop drinking once you had started?

Never

787

80

712

76

Less than monthly

64

6

119

13

Monthly

56

6

63

7

Weekly

55

6

41

4

Daily or almost daily

26

3

8

1

Missing

2

0

4

0

In the last year, have you failed to do what was normally expected of you because of drinking?

Never

895

91

795

84

Less than monthly

53

5

95

10

Monthly

22

2

33

4

Weekly

15

2

16

2

Daily or almost daily

4

0

4

0

Missing

1

0

4

0

In the last year, have you needed a first drink in the morning to get yourself going after a heavy drinking session?

Never

933

94

785

83

Less than monthly

24

2

50

5

Monthly

12

1

36

4

Weekly

16

2

50

5

Daily or almost daily

4

0

22

2

Missing

1

0

4

0

In the last year, have you had a feeling of guilt or remorse after drinking?

Never

649

66

587

62

Less than monthly

161

16

187

20

Monthly

77

8

75

8

Weekly

72

7

82

9

Daily or almost daily

30

3

12

1

Missing

1

0

4

0

In the last year, have you been unable to remember what had happened the night before because you had been drinking?

Never

819

83

787

83

Less than monthly

76

8

95

10

Monthly

46

5

33

4

Weekly

37

4

27

3

Daily or almost daily

10

1

1

0

Missing

2

0

4

0

Composite binary score of “never/ever” responses to questions 2–7 abovea

0

563

57

328

35

1

150

15

188

20

2

94

10

145

15

3

90

9

101

11

4

59

6

94

10

5

20

2

52

6

6

10

1

35

4

Missing

4

0

4

0

Composite frequency score of questions 2–7 aboveb

6

563

57

328

35

7–9

219

22

210

22

10–30

204

21

410

43

Missing

4

0

4

0

aResponses of “never” are coded as 0, any other response is coded as 1, the composite is the cumulative score of binary responses to questions 2–6 above

bA response of “never” received 1 point, “less than monthly” received 2 points, “monthly” received 3 points, “weekly” received 4 points, and “daily or almost daily” received 5 points. The composite scale runs from 6–30, with a score of 6 representing “never” responses to all 6 questions, and a score of 30 representing “daily or almost daily” responses to all 6 questions

Frequencies of the demographic and socioeconomic cofactors among participants reporting the binary alcohol use variable and unprotected sex are presented in Tables 3 and 4. In Rwanda, participant age was not significantly associated with alcohol consumption, but younger men were more likely to have unprotected sex (χ = 2.29, df = 4, P < 0.05). The two largest religions, Catholicism and Pentecostalism, were both strongly associated with alcohol intake—Catholic men were more likely than non-Catholics to drink (90% vs. 56%, χ = 6.53, df = 1, P < 0.05), while Pentecostal men were less likely to drink compared to non-Pentecostals (53% vs. 76% among men, χ = 1.73, df = 1, P < 0.05). Pentecostal men were more likely than non-Pentecostals to have unprotected sex (χ = 7.34, df = 1, P < 0.05). Reporting unprotected sex decreased significantly the longer a couple had been cohabiting (χ = 3.82, df = 1, P < 0.05).
Table 3

Unprotected sex and alcohol consumption by female demographic and socioeconomic factors, Rwanda and Zambia, 2003–2005

 

Rwanda

Zambia

Male drank alcohol in last year (= 702)

Female reports sex without condom at 3-month visit (= 227)

Male drank alcohol in last year (= 669)

Female reports sex without condom at 3-month visit (= 216)

%

n

χa

%

n

χa

%

n

χa

%

N

χa

Age (years)

≤25

74

231

 

27

82

 

73

249

 

40

87

 

26–29

72

144

 

27

52

 

69

128

 

47

49

 

30–33

67

144

 

21

43

 

70

125

 

32

37

 

34–39

70

122

 

21

34

 

72

109

 

31

30

 

≥40

73

55

 

20

15

 

66

57

 

28

13

 

Missing

55

6

1.02

2

1

1.89

100

1

0.64

  

2.91**

Worked for money in past year

All year

70

173

 

22

53

 

68

209

 

32

63

 

Seasonally

70

58

 

23

18

 

73

82

 

51

32

 

Once in a while

74

75

 

28

28

 

69

31

 

43

12

 

Unemployed

71

393

 

24

128

 

72

347

 

37

109

 

Missing

75

3

0.21

0

0

0.52

0

0

0.61

0

0

5.71

Monthly pay (ZMK and FRW converted to USD)

0

71

393

 

24

127

 

72

348

 

37

109

 

$1–15

77

120

 

23

34

 

70

89

 

39

30

 

$16–30

74

125

 

28

44

 

72

102

 

47

40

 

$31–60

61

46

 

25

18

 

65

76

 

24

18

 

>$60

50

18

10.91*

12

4

1.22

70

54

0.66

37

19

5.71

Where lived until 16 years old

City

66

115

 

26

44

 

73

406

 

34

115

 

Town

70

49

 

24

16

 

68

160

 

46

67

 

Village

72

538

3.02

23

167

0.68

66

103

4.24

35

34

4.38**

Religion

Catholic

90

271

 

27

76

 

75

171

 

39

55

 

Non-Catholic

62

431

6.63*b

23

151

3.42b

69

498

4.24b

75

161

0.80b

Pentecostal

62

190

 

26

77

 

65

194

 

37

70

 

Non-Pentecostal

75

512

5.41*b

23

150

1.98b

73

475

6.63*b

37

146

0.02b

aFisher’s Exact Test, unless otherwise specified

bChi-squared

** P-value < 0.05

* P-value < 0.01

Table 4

Unprotected sex and alcohol consumption by male demographic and socioeconomic factors, Rwanda and Zambia, 2003–2005

 

Rwanda

Zambia

Male drank alcohol in last year (= 702)

Female reports sex without condom at 3-month visit (= 227)

Male drank alcohol in last year (= 669)

Female reports sex without condom at 3-month visit (= 216)

%

n

χa

%

n

χa

%

n

χa

%

n

χa

Age (years)

≤25

73

68

 

33

29

 

75

54

 

40

15

 

26–29

75

104

 

31

42

 

73

116

 

50

49

 

30–33

67

137

 

24

47

 

73

155

 

39

51

 

34–39

74

159

 

23

47

 

71

163

 

37

55

 

≥40

69

228

 

19

61

 

66

180

 

28

46

 

Missing

55

6

1.12

2

1

2.29**

100

1

1.41

0

0

1.24**

Worked for money in past year

All year

72

374

 

26

129

 

72

363

 

35

108

 

Seasonally

72

78

 

20

21

 

70

175

 

34

53

 

Once in a while

68

159

 

22

49

 

76

41

 

54

19

 

Unemployed

71

87

 

22

26

 

66

90

 

43

36

 

Missing

57

4

0.62

5

2

1.02

0

0

0.94

0

0

4.75

Monthly pay (ZMK and FRW converted to USD)

0

70

86

 

22

25

 

64

85

 

44

35

 

$1–15

71

92

 

27

34

 

74

31

 

38

10

 

$16–30

71

176

 

26

62

 

74

126

 

39

43

 

$31–60

74

225

 

24

69

 

71

220

 

41

83

 

>$60

66

123

1.31

21

37

0.66

70

207

1.05

27

45

7.30**

Where lived until 16 years old

City

63

112

 

26

44

 

78

352

 

39

108

 

Town

67

58

 

24

21

 

65

165

 

37

59

 

Village

73

532

7.82**

23

163

0.71

63

152

9.21*

34

49

1.19

Religion

Catholic

90

388

 

22

93

 

79

207

 

41

69

 

Non-Catholic

56

314

6.53*b

25

134

1.93b

67

462

9.21*b

36

147

3.21b

Pentecostal

53

110

 

29

60

 

63

149

 

37

55

 

Non-Pentecostal

76

592

1.73*1b

22

167

7.34**b

73

520

9.21*b

37

161

0.27b

Time cohabitating with partner in project

6 months to 2 years

76

192

 

30

73

 

75

179

 

43

54

 

3–5 years

72

206

 

27

74

 

72

198

 

40

70

 

6–9 years

67

149

 

19

41

 

69

115

 

38

40

 

10+ years

68

155

4.63

18

39

3.82*

67

177

2.56

30

52

6.45

aFisher’s Exact Test, unless otherwise specified

bChi-squared

** P-value < 0.05

* P-value < 0.01

Broadly, similar trends were seen in Zambian couples. Both younger men (χ = 1.24, df = 4, P < 0.05) and younger women (χ = 2.91, df = 4, P < 0.05) were more likely to have unprotected sex. In Lusaka, though, women from mid-sized towns were more likely to report unprotected sex (χ = 4.38, df = 2, P < 0.05). The same trends in drinking were also seen among Catholics and Pentecostals as in Rwanda, but there were no associations between religion and unprotected sex.

In multivariate modeling for Rwanda (Table 5), the two composite measures of alcohol consumption were not significantly associated with unprotected sex, but the binary variable describing alcohol use in the year prior to enrollment was significant. In the final, fully-adjusted model, only three variables were associated with unprotected sex. Males who reported alcohol use in the year prior to enrollment were 1.47 times as likely to have unprotected sex as their non-drinking counterparts (OR 1.47, 95%CI: 1.02–2.11, P < 0.05). Compared to other religions, male Pentecostals were more likely to report unprotected sex (OR 1.62, 95%CI: 1.12–2.33, P < 0.05). Relative to those who lived together for 2 years or less, couples who lived together for 6–9 years and those who lived together for more than 10 years were also less likely to report having unprotected sex (OR 0.54 95% CI: 0.34–0.84, P < 0.05 and OR 0.53, 95% CI: 0.34–0.83, P < 0.05 respectively).
Table 5

Regression analysis of female-reported unprotected sex at 3-month follow-up visit; Rwanda and Zambia, 2003–2005a (N = 921)

 

Rwanda

Zambia

Odds Ratio

95% Confidence Interval

Odds Ratio

95% Confidence Interval

Male reports alcohol use in year prior to testing/enrollment

1.47

(1.02–2.11)

1.67

(1.13–2.48)

Male is of Pentecostal faith

1.62

(1.12–2.33)

Time cohabiting with partner in study

6 months to 2 years (ref)

3–5 years

0.88

(0.60–1.30)

6–9 years

0.54

(0.34–0.84)

10+ years

0.53

(0.34–0.83)

Electricity in household

1.87

(1.23–2.83)

Where female lived until 16 years old

City (ref)

Town

1.66

(1.11–2.48)

Village

1.07

(0.66–1.74)

Composite binary score of “never/ever” responses to 6 questions on alcohol consumption

0

1

1.01

(0.65, 1.37)

1.02

(0.61, 1.43)

2

0.98

(0.58, 1.28)

1.01

(0.42, 1.61)

3

1.02

(0.70, 1.34)

0.97

(0.32, 1.62)

4

1.03

(0.50, 1.56)

0.98

(0.23, 1.73)

5

0.97

(0.46, 1.48)

1.03

(0.42, 1.64)

6

0.99

(0.42, 1.56)

1.02

(0.46, 1.58)

Missing

1.02

(0.46, 1.58)

0.96

(0.43, 1.39)

Composite frequency score of 6 questions on alcohol consumption

6

7–9

1.02

(0.42, 1.62)

0.99

(0.23, 1.75)

10–30

0.98

(0.53, 1.43)

0.98

(0.52, 1.50)

Missing

1.03

(0.36, 1.70)

1.01

(0.37, 1.65)

 

0.99

(0.66, 1.32)

1.04

(0.49, 1.59)

aModels also controls for age difference between the couple, male and female amount of time working in the past year, couples’ combined monthly income, number of years lived in Kigali for male and female, setting (city, village, town) where male and female participants lived until age 16, female being Pentecostal, male and female reading and understanding French, whether the couple has a legally or civilly recognized marriage, whether the couple owns the home they live in, and whether the couple’s home has electricity, and whether the couple wish to have more children as well as interaction variables between male drinking in the previous year and male and female Pentecostal faith. All of the above variables failed to reach levels of statistical significance

In Zambia, the composite alcohol consumption measures were not significant. In the final model for Zambia (Table 5), three variables were associated with unprotected sex. Men who reported alcohol use in the year prior to enrollment were 1.67 times as likely to have sex without a condom at least once in the three months following enrollment (OR 1.67, 95%CI: 1.13–2.48, P < 0.05). Women who grew up in towns were significantly more likely to report unprotected sex (OR 1.66, 95% CI: 1.11–2.48, P < 0.05) than women who grew up in cities, and couples living in homes without electricity were nearly twice as likely to report unprotected sex (OR 1.87, 95%CI: 1.23–2.83, P < 0.05).

Discussion

The demographic and socioeconomic profile of study participants in both Rwanda and Zambia showed a largely poor urban population. Alcohol use was highly prevalent among men, with binge drinking more prevalent in Zambia, but less than half of women reported any alcohol use in the year prior to enrollment, which shifted our focus to analyzing male alcohol consumption. Of all couples enrolled, 23% reported unprotected sex at their first follow-up visit. It is important to recognize that the cohort analyzed in this study is a highly select group attending a clinic-based research facility. Despite some differences in baseline demographic descriptors and alcohol consumption patterns and despite the differences between Rwanda and Zambia in general, the overall results remained the same; male alcohol use was associated with unprotected sex in sero-discordant couples.

However, it is interesting that only one measure of alcohol consumption was significantly associated with unprotected sex: the remaining two alcohol variables which were intended to capture the severity of alcohol consumption were not significantly associated with unprotected sex. This surprising result points to the potential under-reporting of alcohol behaviors by study participants. It is suggested that while men may be willing to report alcohol consumption, they may under-report negative behaviors such as loss of memory or feelings of guilt due to alcohol (Table 2 shows that the majority of respondents answer negatively to the AUDIT questions). Alternatively, the AUDIT questions may not be culturally appropriate for measuring alcohol behaviors in African settings; further qualitative research is required to assess how the AUDIT questions are interpreted in these settings.

Couples in the analysis had been through at least two separate intensive counseling and intervention sessions, the first at their CVCT session, the second at enrollment into the study, designed to educate them about their HIV sero-discordant status and to help them begin to protect their own health, and the health of their partners. After controlling for many other risk factors, very few variables remained significantly related to unprotected sex and the only one that is modifiable is alcohol use.

Many mechanisms have been proposed to explain alcohol’s relationship to the transmission of HIV, some biological and some behavioral. The data presented are suggestive of a behavioral link. While the exposure was analyzed in a historical context, i.e., drinking in the year prior to enrollment, it may be assumed that most of the men who reported alcohol use in the past year continued to use alcohol during the three-month follow-up interval. Whether their intoxication, however slight, led to increased desire for unprotected sex, often seen as more pleasurable, or a minimization of the perceived risk or infection that unprotected sex presented to themselves and their partners, it seems plausible that alcohol consumption leads to unprotected sex. While this cross-sectional study is suggestive of a relationship between alcohol consumption and unprotected sex, further research is necessary to determine whether alcohol consumption is linked to sero-conversion.

The loss of nearly 40% of enrolled couples to follow-up in Zambia is a major limitation of the analysis, but it is reassuring that, even with a high rate of loss to follow-up, the Zambian analysis with regard to alcohol consumption was very similar to the Rwandan analysis. A comparison of those lost to follow-up with the remaining study participants showed no significant differences in background demographic, behavioral or socio-economic characteristics, suggesting there is no systematic bias. Another limitation is that there is no explicit causal explanation for the relationship between alcohol use and unprotected sex; however this is not possible with cross-sectional data. At the time of the study ART were not commonly available in either Rwanda or Zambia, hence the present analysis does not control for the influence of ART on the perception of health, and how this may influence the perceived need for condom use. Additional limitations include the potential for social desirability to influence the reporting of both unprotected sex and alcohol consumption, and the temporal disparity of associating alcohol use in the previous year with unprotected sex measured for the past 3 months. However, despite these limitations, we find an association between alcohol consumption and unprotected sex; further research is needed to examine the relationship between ART, alcohol consumption and condom use, and longitudinal designs should be employed to disentangle to temporal ordering of alcohol and unprotected sex.

Conclusion

HIV-negative partners in sero-discordant couples are at high risk of acquiring HIV infection. Given the demonstrated relationship between alcohol consumption and unprotected sex, when a couple is found to be sero-discordant counselors should ask the couple questions about their alcohol consumption, and focus discussion on the potential influences of alcohol use on condom use. Given that HIV acquisition has been shown to increase when alcohol is consumed before sex (Zablotska et al. 2006), this approach is reasonable. By making couples aware that alcohol use may inhibit condom use, the rate of HIV transmission could decrease. This finding is not limited to the setting of CVCT, however, as the same principles apply to individual-level VCT, a more common practice in sub-Saharan Africa. Messages on the potential inhibiting effect of alcohol consumption on condom use need to be routinely included in all behavioral change interventions. This is particularly important in contexts, such as many Sub-Saharan African countries, where high rates of alcohol consumption and HIV prevalence co-exist.

Copyright information

© Springer Science+Business Media, LLC 2007